Molecular represention in AI-assisted synthesis chemistry: from sequence to graph

Abstract

<p indent="0mm">With the rapid development of deep neural networks and their derived algorithms, artificial intelligence (AI) has flourished in various fields in recent years. In the field of computer-aided synthetic route design, artificial intelligence algorithms such as generative adversarial networks and variational autoencoders have performed very well. How to deal with the geometric representation of molecules is a key step in transforming chemical problems into algorithmic problems. Starting from different dimensions, this paper systematically summarizes the application of various molecular sequence models, graph theory and other representation methods in AI-assisted synthesis chemistry, and discusses the shortcomings and challenges of current molecular representation.